TAR based shape features in unconstrained handwritten digit recognition

Size: px
Start display at page:

Download "TAR based shape features in unconstrained handwritten digit recognition"

Transcription

1 TAR based shape features n unonstraned handwrtten dgt reognton P. AHAMED AND YOUSEF AL-OHALI Department of Computer Sene Kng Saud Unversty P.O.B. 578, Ryadh 543 SAUDI ARABIA shamapervez@gmal.om, yousef@s.edu.sa Abstrat: - In ths researh, the reognton auray of trangle-area representaton (TAR based shape feature s measured n reognzng the totally unonstraned handwrtten dgts. The TAR features for dfferent trangles of varable sde lengths that are formed by takng the ombnatons of dfferent ontour ponts were omputed. The set of ontour ponts that yelded the best features was expermentally dsovered. For lassfaton a urve mathng tehnque s used. Several experments were onduted on real-lfe sample data that was olleted from postal zp odes wrtten by mal wrters. The hghest reognton result of 98.5 % was aheved on the tranng data set and 98.3% on the test data set. Key-Words: - Trangle area representaton, Shape desrptors, Dgt reognton, Contour ponts, and zp odes Introduton One of the advantages of shape desrptor based features s that they provde strutural desrpton of the underlyng pattern for reognton. Ths has attrated researher s attenton []- [4]. As a result, a large number of shape desrpton tehnques have evolved. These tehnques are broadly lassfed nto ontour-based and regon-based tehnques. If a tehnque extrats shape features from the ontour (boundary only then t s lassfed as a ontourbased tehnque, and f a tehnque extrats the property of a regon lke the area overed by a regon then t s lassfed as the regon based tehnque. These tehnques an extrat strutural and global features. A ontour based tehnque an extrat strutural features lke han ode, polygon, B-splne and nvarants whle a regon-based tehnque an extrat strutural features lke onvex hull and meda axs and the lke. Smlarly, global features lke permeter, ompatness, eentrty, shape sgnature, Fourer desrptors an be extrated from the ontour and global features lke area and moments an be extrated from the regons []. In ths paper, the performane of a ontourbased tehnque referred to as trangle-area representaton (TAR sgnature has been measured n reognzng the totally unonstraned handwrtten postal zp ode dgts [5] and [6]. A TAR sgnature reflets ontour haratersts lke onavty or onvexty at ontour ponts at whh t s omputed [][3] [4] and [5]. In addton, t s very effetve n apturng the loal as well as global shape haratersts by smply varyng the trangle sde-lengths. Its man advantage s that t an be used to defne translaton, rotaton, and sale nvarant shape features (also referred to as shape desrptors n ths paper. There are many varatons of TAR sgnatures. In ther paper Roh and Kweon [] have defned TAR based shape features usng fve equally spaed ontour ponts P (t, P (t, P 3 (t, P 4 (t and P 5 (t form a lst of N ontour ponts where the pont P (t s the o-ordnate (x (t, y (t of the th ontour pont, and t denotes the set of ontour ponts seleted at tme t. For eah seleton t=,,3,,n they defned the shape nvarant P5 P P4 P5 P P3 I(t as: I =, ISSN: Issue 5, Volume 9, May

2 where, P t P (, P t P (, 5( 4 t 5( 4 t 5( 4 t P t P ( and P t P ( are the areas 5( 4 t of trangles formed by jonng the ponts: {(P5(t, P(t, P4(t, (P5(t P(tP3(t, (P5(t, P(t, P3(t and (P5(t P(t P4(t}, and the fve ontour ponts were determned as: to as the reognton module. Ths module aepts the lst of the predted lasses and determnes the atual dentty of the unknown sample wthn these lasses. P (t=(x (t, y (t, P (t= (x (N/5+t, y (N/5+t, P 3 (t= (x (N/5+t, y (N/5+t, P 4 (t= (x (3N/5+t, y (3N/5+t, P 5 (t= (x (4N/5+t, y (4N/5+t. They obtaned the shape sgnature of a pattern by plottng the value I(t versus t for dfferent values of t=,,3,, N, and tested the performane ths desrptor n mage retreval. Through a seres of experments, they have demonstrated that these features are nvarant under rotaton, translaton and sale, and also stable and robust. In a smlar approah, El Rube et. al [5] have tested the robustness of a mult-sale trangle-area representaton for -D shapes and dsovered that the representaton s least affeted by nose. In further experments, Alajlana et. al [3] and [4] have dsovered that the TAR based shape desrptors are adaptve as they an apture the varyng levels of detals smply by ontrollng the seleton of the number of ontour ponts whle generatng the sgnature. These propertes of the TAR sgnature based shape desrptors are the man motvaton to test ther performane n terms of reognton auray on real-lfe samples of unonstraned handwrtten dgts. For ths purpose, the test data set that s desrbed n Ahmed and Suen [4] and Abuhaba and Ahmed and [5] s used. Ths data set reflets natural varatons and dstortons n shape, sze and orentaton. The data was obtaned from the totally unonstraned handwrtten postal zp odes that were olleted by the US postal serve department from the dead letter envelopes at offes aross the ountry and they were dgtzed on 6 gray levels. In the experments, reported n here, samples possessng almost all the natural varatons exept breaks n shapes were used. Fg. below shows representatve samples from the data set. In searh of a better reognton result, stagewse lassfaton shemes were developed n whh the frst stage s the predton stage that s referred to as predton module hereafter. Ths module was desgned to predt the probable lasses of an unknown sample. The seond stage s referred Fg. Data Set Representatve Samples The predton module uses a zone-based predton method. In ths method, haratersts of the jont dstrbuton of two TAR feature elements were studed to determne a vable zone sze. The predton helped n narrowng down the dentty of the unknown patterns. As mentoned before, the reognton stage reognzes the pattern as one of the member of the predted pattern lasses. In ths stage a urve mathng tehnque s used. The urve mathng s an atve area of researh and ts applatons are beng explored n searhng shapes where shapes are represented as parameterzed urve n twodmensons [7]-[4]. Ths representaton redues the shape mathng problem to the problem of mathng two urves n a two-dmensonal Euldean spae, and urve mathng tehnques, lke dynam spae warpng desrbed n [3], and Fr ehet dstane [8]-],[] and [4] whh s a natural measure for mathng urves, are beng devsed and tested. We have developed a new urve mathng tehnque that uses the umulatve feature urve obtaned from the unknown pattern and mathes t aganst a set of prototypal urves representng the average umulatve feature urves of the known pattern lasses. The prototypal urves are feature growth urves that are estmated from the tranng set samples. ISSN: Issue 5, Volume 9, May

3 The stage-wse lassfaton sheme has yelded a very promsng orret reognton result of 98.5 % and 98.3% on the tranng and test set data respetvely. The TAR tehnque s presented n Seton. The desgn of the predton stage s desrbed n Seton 3 and the reognton stage n Seton 4. The expermental results are represented n Seton 5 and a dsusson on the performane n Seton 6. Trangular-Area Representaton (TAR The Trangular Area Representaton (TAR s omputed from a rular lst of ontour ponts. The lst s obtaned from the pattern mage I[M,N] havng M rows and N olumns. The pont I[,] s the top leftmost pont and I[M-, N-] s the bottom rghtmost pont. Let the lst of ontour ponts be P (x, y, P (x, y,, P n- (x n-, y n-, where n s the total number of ontour ponts. The pont P (x, y, denotes the x and y o-ordnate of the th ontour pont P. The pont P (x, y s the fst ontour pont that s deteted by sannng the mage row-by-row startng from the pont I[,]. One the frst boundary pont s deteted, startng from that pont the rest of the ontour ponts were deteted and reorded by followng the ontour lokwse. The ontour followng proess stops at the last ontour pont P n- (x n-, y n-, whh s n the ontour pont n the 8-neghborhood of the start pont P (x, y. To ompute a TAR value at the pont P (x, y, three ontour ponts from the lst, say P k (x k, y k, P (x, y and P r (x r, y r, where k < < r were seleted and the area of the trangle formed by these ponts was omputed by evaluatng the determnant value A as shown n equaton. It an be easly verfed that the value of A s whenever (x k, y k, (x, y and (x r, y r are o-lnear, and A < for onvex and A > for onave regons (Alajlana et. al, 7. The sgned value of A provdes an estmate of the dstrbuton of the three ontour ponts (x k, y k, (x, y and (x r, y r. xk yk A = x y ( x y r A TAR sgnature s a plot of trangle number versus the area of the trangle. For example, f P k (x k, y k, P (x, y and P r (x r, y r, where k < < r, and r =,,, m, are a set of m trangles, then the plot of (, A for =,,, m s the TAR sgnature. The TAR sgnature of the mage of a handwrtten dgt zero s shown n Fg.. Ths sgnature was obtaned by trangles formed by three onseutve ontour ponts (x -, y -, (x, y, (x +, y + for =,, n-, where n =65 s the total number of ontour ponts. The sgnature s normalzed to one by dvdng eah TAR value A by M N (M=3 and N=9. TAR Sgnature Dgt Zero Fg. TAR sgnature of the Dgt Zero Image 3 Predton Module Desgn The two best TAR features (the features that yelded the maxmum orret reognton were expermentally determned for predton module desgn. Let these features be denoted as f and f respetvely. In mplementaton, these features are the TAR values of trangles P P P and P P 3 P as shown n Fg 3 (a. The ponts P and P are the start and mdpont n the ontour and ther oordnates are (x,y and (x n/,y n/ respetvely, where and n/ are ther ndes n the ontour pont lst. Smlarly, the ponts P and P 3 are at the ¼ rth and ¾ rth postons of the ontour pont lst and ther oordnates are (x n/4,y n/4 and (x 3n/4,y 3n/4 respetvely, where n/4 and 3n/4 are the ndes of these ponts n the ontour pont lst. To predt the probable lasses of an unknown pattern, the jont dstrbuton of the feature values of f and f n two dmensonal feature spae was estmated from the tranng set samples, and the predton zones were estmated. Here a zone s defned as a retangular regon. The jont dstrbuton of f and f for tranng set samples of eah pattern lass to 9 s shown below n Fg. 3 (b. In the fgure, the values of f are used for horzontal axs and the values of f for vertal axs. Eah feature par (f, f n the plot represents a pattern lass. ISSN: Issue 5, Volume 9, May

4 The zones were formed by parttonng the feature axes f and f nto n equal parts. If these 3 n parts are denoted as f, f, f,..., f and 3 n f, f, f3,..., f respetvely, then zones are j j defned as retangles {( f, f bl,( f, f tr }, for,j =,,3,, n, where, n s the number of horzontal j and vertal parttons, and ( f, f bl and j ( f, f tr are the bottom-left and top-rght ponts of the retangle. The zones are reorded as a tuple k k k k Z k ={ ( f, f bl, ( f, f tr, L k }, where the frst two felds are the bottom-left and top-rght orner ponts that defne the zone and L k s a lst that ontans labels of those sample patterns whose feature values f and f are lyng n zone Z k f (TAR P 3 P Fg. 3 (a Seleted ontour ponts Feature & Jont Dstrbuton f (TAR Fg.3 (b: Jont dstrbuton of features f and f n dgts,,,, 9 The zone wdth and heght were estmated from the tranng set samples by estmatng an optmal partton sze that yelded the best zone formaton P P.e., no zone ontaned more than a pre-spefed number of dstnt pattern lasses. After estmaton, all the non-empty zones were reorded as a tuple n an array Z of tuples [Z Z Z 3,, Z m ], where m s the total number of non-empty zones n n n retangles. To predt the lasses for a gven feature par (f,f the array Z s searhed to loate the zone n whh the feature par les. If t les, say n zone Z k, then the predted lasses are the labels lst L k that belongs to Z k. 4 Reognton Module Desgn Ths module uses TAR features but these features are dfferent from those two features that are used n the predton module. These features are extrated from the TAR sgnatures that are defned to apture more boundary detals whh are obtaned by takng more than two trangles of varyng sde lengths at dfferent ontour loatons. The extraton of these features s desrbed n Seton 4.. For lassfaton, the prototypes that represented the lasses were generated for the TAR sgnatures of eah lass samples. The prototype generaton proess s desrbed n Seton 4.. To obtan a better reognton sore a urve mathng based lassfaton proedure was developed whh s gven n Seton Feature Extraton As mentoned before, trangles havng sdes of dfferent lengths and are loated at dfferent ontour postons were used for TAR sgnature generaton. One of the advantages of ths approah s that t gves flexblty n devsng exploratory feature extraton tehnques. Consequently, t helps n determnng an optmal dsrmnatory TAR feature set from the ombnatons of trangles havng sdes of dfferent lengths and they are loated at dfferent ontour postons. In searh of suh a feature set for unonstraned dgt reognton several experments were onduted and the defnton and extraton of the feature set that yelded better reognton result s desrbed below. In ths ase, the feature set was obtaned from the ontour part P to P n/3 n n pont ontour lst P, P,, P n-. Ths ontour part was parttoned nto ℵ equal segments. By hoosng dfferent values of ℵ, dfferent feature sets an be formed. Assume that the endponts of eah of the ℵ segments are: (P (k- τ, P k τ, where k=,,3,, ℵ, are segment numbers and τ= n/(3 ℵ s the segment length. For the segment number (k= the segment endponts are P and Pτ, where P s the ontour start pont and Pτ s the pont after τ ponts from the start pont. ISSN: Issue 5, Volume 9, May

5 Smlarly, for k=ℵ the segment endponts are P n/3- τ and P n/3. For eah k=,,3,, ℵ a trangle wth three ontour ponts (P (k- τ, P n/3+(k- τ, P n/3+(k- τ an be formed and ts TAR value s onsdered as the feature f k. The feature values are normalzed to. For larty onsder Fg. 4 and assume that t ontans 45 ontour ponts (n=45. In ths fgure, the ontour part P P 5 s dvded nto 3 (ℵ=3 segments: P P 5, P 5 P and P P 5 where the segment length s τ =5 ponts. In ths ase, three features f, f and f 3 an be generated by omputng the areas of the trangles (P P 5 P 3, (P 5 P P 35 and (P P 5 P 45 respetvely. Values Average Feature Values Features Fg.5: Plot of Average Feature Values Cumulatve Average Feature Values P 7 P 35 P 3 P 4 P P 5 Cumulatve values Features P 5 Fg. 4 TAR Sgnature Computaton Fg.5 shows a plot of the average of eah feature value of 6 (ℵ=6 features observed n sample wth samples per dgt. In the plot, dgt lasses are represented by dfferent olors (see the sde legend. To study the ontrbuton of ndvdual feature n the aumulatve growth of feature values, features f k, k=,,, ℵ, were transformed nto a growth funton form: P k s( k = f, where s(k s j= an nreasng funton as s(k s(k+ for k =,,, ℵ. The advantage of ths representaton s that t gves smple representaton (see Fg. 6 as ompared to the orgnal sgnature (Fg.5. j P 5 Fg.6: The Cumulatve Average Feature Values 4. Prototype Creaton The prototypes are lass representatves. In ths researh prototypes (one for eah lass were reated from the average umulatve feature values s ( k for k=,,, ℵ and for lass =,,, m, where m s the number of lasses and ℵ s the number of features. The values of s ( k are N omputed as s ( k = s ( k / N, where (k :s = the umulatve value of the k th feature of the th sample n lass, and N s the total number of samples n lass. As mentoned before, n ths researh the average feature values s ( k are used as prototype. One of the advantages of usng the umulatve feature values s that the prototypes an be modeled usng known growth funtons. In that ase, nstead of storng the average values, only the model parameters of the growth funton need to be estmated, and the prototypes an be generated n real-tme and prototype urve shapes an be vared s ISSN: Issue 5, Volume 9, May

6 n real-tme to aommodate the varatons; otherwse to apture the varatons several prototypes may be requred for eah lass. 4.3 Classfaton The lassfaton module aepts the umulatve feature urve s(k of an unknown pattern, where k=,,, ℵ and ℵ s the total number of features. The lassfer lassfes s(k by omparng t aganst the prototypes s ( k, k=,,, ℵ of eah lass =,,, m. The lassfaton rtera s to lassfy s(k nto lass, f the value of the funton Ψ (s(k, s ( k s maxmum for =,,, m, where funton Ψ (s(k, s ( k ompares the two urves s(k and s ( k and yelds a smlarty ndex. The step-wse urve omparng proess s as follows.. Create a feature-wse dfferene table δ between the two urves s(k and s ( k by k, omputng the valueδ = s( k s ( k, for k=,,, ℵ and =,,, m.. Create a rank table γ k, by assgnng a rank to the elements of eah row of δ k, by the followng rules: γ k, =, f δ k, s the smallest value, γ k, =, f δ k, s the next hgher value and so on.. Fnally, γ k, = m f δ k, s the hghest value. Note: In the rankng proess a te may our. In whh ase, assgn average rank to all the elements that are n te. For example, f δ, δ and δ k, k, k, k, 3 have equal values for the lasses, and 3, then the ranks for γ = γ = γ = ( + + 3, /3. k, k, k, 3 C. Compute ℵ d = γ k, k= The d value an also be used to lassfy an unknown pattern nto lass f d s the mnmum of d, =,, 3,, m and. Experments were onduted wth ths lassfaton proedure but the result shows poor reognton performane (see Seton 5. To mprove the reognton performane a weghted smlarty measure, desrbed below, was developed and used. A. Compute the rank frequeny table R for eah lass =,,3,, m usng the tranng set samples. In ths table, eah entry R =( γ s k, the frequeny wth whh the k th feature ranks the tranng sample of some th lass as the member of lass. B. For an unknown pattern obtan γ k, as desrbed n steps and before. C. For eah lass =,,3,, m ompute γ k, φ k, = for all k =,,3,, ℵ and = γ k,,,3,, m. D. Compute the funton = ℵ k= ψ ( s( k, s ( k φ. k, E. Classfy an unknown pattern havng umulatve feature values s(k nto lass f Ψ (s(k, s ( k s the maxmum for the lass. Ths lassfaton proedure mproved the reognton auray onsderably (Seton 5 5 Experments As mentoned before, several experments were onduted n searh of the best possble reognton sore. Ths seton desrbes these experments along wth ther respetve feature extraton and lassfaton methods, and reognton performane. Experment 5 s the man experment. The other experments are desrbed just to llustrate the performane of the tehnques that provded nsght that has lead to the development of mproved tehnques. 5. Experment The objetve of the frst experment was to assess the applablty of TAR feature n unonstraned handwrtten dgt reognton. In ths experment the TAR feature of a sngle trangle of sde length n/3, where n s the total number of ontour ponts was used. The trangle was formed by seletng the ontour ponts (P -n/3, P, P +n/3 for a gven [n/3, n/3] and the TAR feature was omputed (as desrbed n Fg 4 Seton 4.. Durng the tranng phase, for every seleted trangle the average feature value f and ts varane σ for eah lass were estmated. Usng these values an unknown pattern was lassfed as omng from the lass f ts TAR value belongs to the nterval f ± kσ where k s a hosen onstant. In ISSN: Issue 5, Volume 9, May

7 ase, the TAR value does not belong to any lass nterval then the unknown pattern s delared unreognzed. Several experments were onduted usng dfferent trangle loatons and the reognton performane for dfferent TAR values was estmated. The best orret reognton performane of.% was observed on tranng set samples for k=.. 5. Experment In ths experment the jont dstrbuton of two TAR feature values that were obtaned from the two trangles was studed. Lke experment the sde length of both the trangles were taken as n/3 and they were seleted from two dfferent loatons n the range as desrbed n experment. From the tranng set samples, the lass feature means for eah varanesσ and f and f, and lass feature σ were estmated and the,, ntervals of the form( f kσ,, f ± kσ, ± were formed for eah lass, where k and k are the hosen onstants. Ther values were estmated from the tranng set samples. In ths ase, an unknown pattern havng feature values f and f was reognzed as omng from the lass f the feature values f and f fall n the lass nterval( f ± kσ,, f ± kσ,. Experments were onduted for several trangle ombnatons and for eah ombnaton k and k values were hosen. In these experments the trangles formed by the ponts P, P, P and P 3 shown n Fg.3 (a yelded the best reognton result. The horzontal trangle ombnatons ( P 3 P P and P P P 3 yelded 4.% and the vertal trangle ombnatons ( P P P and P P 3 P yelded 43.7% orret reognton on the tranng samples that were used n experment. 5.3 Experment 3 In ths experment, features obtaned from the horzontal and vertal sets of trangles, desrbed n experment, were used but n two dfferent lassfaton shemes. Both the shemes use predton and reognton modules. To lassfy a pattern, frst ts features were presented to the predton module. The output of ths module s a set of probable lasses whh s sent to the reognton module along wth the feature elements. The reognton module reognzes the unknown pattern as one of the members of the predted lasses. In the frst sheme, the TAR features obtaned from the vertal set of trangles were used to reate the zone-based predton module and hene to predt the probable lasses. The TAR features obtaned from the horzontal set of trangles were used to reognze an unknown pattern as one of the members of the predted lasses usng the nterval method desrbed n experment. Ths method mproved the orret reognton perentage to 49.8%. The seond sheme was smlar to the frst sheme but n ths ase the horzontal set of trangles was used to desgn the predton module and vertal set of trangles to test the reognton performane. In ths ase, the reognton performane was 5.3% whh s slghtly better. The experment 3 was onduted on the same tranng set samples that were used n earler experments. 5.4 Experment 4 In ths experment, several ombnatons of two TAR feature values were studed to mprove the predton performane, and multple TAR features representng 4, 8, and 6 features were studed to mprove the reognton performane. Lke experment 3, predton zones were estmated and t was observed that the two vertal trangles that are shown Fg.3 (a yelded the best predton zones n whh no zone ontans more than three predton lasses. Usng samples we observed that the zone heght=.5 and wdth =.5 yelded the best predton zones. To reognze an unknown pattern, a lassfer usng d = K k= γ as desrbed n Seton 4.3 was k, used. It was observed that 6 TAR feature set produed the best orret reognton perentage of 63.4% on the same tranng samples that were used n experments to Experment 5 Ths experment was desgned usng the predton module and the TAR feature set of experment 4, and the weghted rank lassfer desrbed n Seton 4.3. Ths lassfer lassfes an unknown pattern havng feature values s(k as the member of one of the predated lasses : f Ψ (s(k, s ( k s maxmum for that lass. ISSN: Issue 5, Volume 9, May

8 Ths lassfaton proedure mproved the reognton auray sgnfantly to 98.5% that was aheved for 6 features on tranng samples. The onfuson matrx of ths experment s shown n Table I, whh shows that the hghest onfuson ours between dgts 4 and 8 and the next hghest onfuson between 9 and 8. It looks lke strutural varatons n dgts 4 and 8, and 9 and 8 may have yelded almost smlar shapes. The reognton auray was tested on a larger test sample of sze 5 dgts, and t was observed that the auray dropped to 85.%. The onfuson table of ths test s shown n Table II. sde boundary are ommon, whle other ontour ponts n dgt one may have not ontrbuted sgnfant dsrmnatory features. Tranng set sze mght have ontrbuted to the error. So, to measure the effet of the larger tranng set sze, the 5 test samples were used as tranng set, and sample that were used n tranng set were used as the test set. The onfuson of table of the tranng set and test set are shown n Table-III & IV. The reognton perentage on the tranng set s 96.6% and test set s 98.3%. The tranng set perentage s low beause of the sze and shape varatons n dgt one. Table: I. Confuson Table Tranng Set ( Samples Table-III: Confuson Table Tranng Set (5 Samples Reognzed As Reognzed As Unknown dgt Unknown dgt Table-II: Confuson Table Test Set (5 Samples Table-IV: Confuson Table Test Set ( Test samples Reognzed As Reognzed As Unknown dgt Table II shows that the dgt has the lowest reognton perentage of 54%, and most of the errors our between dgts and 4, and 7 and and 8. To nvestgate the reasons, samples of dgt one were examned, and t was dsovered that the sze and rotaton varatons are lkely reasons beause the rotated dgt and dgts 4, 7 and 8 have majorty of ontour ponts that are lyng at the rght Unknown dgt Conluson Ths researh started wth a smple experment that was desgned to measure the effetveness of the TAR features n reognzng the totally unonstraned handwrtten dgts. Intally, a lttle suess was aheved but t was observed that TAR features an be defned n many ways. Thus, t an ISSN: Issue 5, Volume 9, May

9 be used as an exploratory tehnque to explore the best possble ombnaton of TAR features. In the proess, a seres of experments were onduted and every experment yelded an enouragng result that reahed lose to 98.5% orret reognton sore whh s an exellent sore n the lght of the data qualty of the test and tranng sets. The data was obtaned from the totally unonstraned handwrtten postal zp odes. These zp odes were olleted by the US postal serve department from the dead letter envelops at offes aross the Unted Sates of Amera. The data reflets expeted varatons n wrtng styles, stroke thkness, wrtng materal, nk other smlar attrbutes. The zp-odes were dgtzed on the 6 gray levels usng very rudmentary sannng tehnology that produed some struturally damaged mages. The TAR features are strutural features. Extraton of these features from the broken mages may not reflet the true stuaton. Therefore ther performane on the struturally damaged samples s expeted to deterorate. However, the expermental results show that on the struturally undamaged samples havng almost all the natural varatons exept the breaks n shapes, the tehnque yelded exellent results. Referenes: [] D. Zhang and G. Lu, Revew of shape representaton and desrpton tehnques, Pattern Reognton, Vol. 37, 4, pp. 9. [] K. Roh and I. Kweon, -d objet reognton usng nvarant ontour desrptor and projetve refnement, Pattern Reognton Vol. 3 No. 4, 998, pp [3] N. Alajlana et al., Shape retreval usng tranglearea representaton and dynam spae warpng, Pattern Reognton Vol. 4, 7, pp [4] Naf Alajlan, Mohamed S. Kamel and George H. Freeman, Geometry-Based Image Retreval n Bnary Image Databases, IEEE Transatons on Pattern analyss and Mahne Intellgene, Vol. 3, No. 6, 8, pp. 3-. [5] El Rube, N. Alajlan, M. Kamel, M. Ahmed, G. Freeman, Robust multsale trangle-area representaton for d shapes, n: IEEE, Internatonal Conferene on Image Proessng (ICIP, Genoa, Italy, September 5, pp [6] S. Belonge, J. Malk, and J. Puzha, Shape mathng and objet reognton usng shape ontexts., IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol. 4, No. 4,, pp [7] G. Chauang and C. Kuo, Wavelet desrptor of planar urves: Theory and applatons, IEEE Transaton on Image Proessng, Vol. 5, No., 996, pp [8] E. Klassen, A. Srvastava,W. Mo, and S. Josh, Analyss of planar shapes usng geodes paths on shape spaes., IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol. 6, No. 3, 4, pp [9] F. Mokhtaran and A. Makworth, Sale-based desrpton and reognton of planar urves and two dmensonal shapes, IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol. 8, No., 986, pp [] F. Mokhtaran and A. Makworth, A theory of mult-sale, urvature-based shape representaton for planar urves, IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol. 4 No. 8, 99, [] E.G.M. Petraks, A. Dplaros, and E. Mlos, Mathng and retreval of dstorted and oluded shapes usng dynam programmng, IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol.4, No., pp [] T. Sebastan, P. Klen, and B. Kma, Reognton of shapes by edtng ther shok graphs., IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol. 6, No. 5, 4, [3] Chunjng Xu, Janzhuang Lu, and Xaoou Tang, D Shape Mathng by Contour Flexblty, IEEE Transatons on Pattern Analyss and Mahne Intellgene, Vol. 3, No., 9, pp [4] T. Adamek and N.E. O Connor, A Multsale Representaton Method for Nonrgd Shapes wth a Sngle Closed Contour, IEEE Trans. Cruts and Systems for Vdeo Tehnology, Vol. 4, No. 5, 4, pp [5] P. Ahmed and C. Y. Suen, Computer Reognton of Totally Unonstraned Handwrtten Zp Codes, Int. J. of Pattern Reognton and Artfal Intellgene, Vol. No., 987, pp. -5. [6] I. S. I. Abuhaba and P. Ahmed, "A fuzzy Graph Theoret Approah to Reognze the Totally Unonstraned Handwrtten Numerals," Pattern Reognton, Vol. 6, No. 9, 993, pp ISSN: Issue 5, Volume 9, May

10 35. [7] P. F. Felzenszwalb and J. Shwartz, Herarhal Mathng of Deformable Shapes, Pro. IEEE Conf. Computer Vson and Pattern Reognton, Vol., 7, pp. -8. [8] Kevn Buhn, Make Buhn, Carola Wenk, Computng the Fr ehet Dstane Between Smple Polygons, nd European Workshop on Computatonal Geometry, 6, pp. 3-6 [9] H. Alt and M. Godau, Computng the Fr ehet dstane between two polygonal urves, Internat. J. Comput. Geom. Appl., Vol. 5, 995, pp [] Adran Dumtresu and G unter Rote, On the Fr ehet dstane of a set of urves, 6th Canadan Conferene on Computatonal Geometry,Montreal, Quebe, 4, pp [] Xao-Dao Chen, Lnqang Chen, Ygang Wanga, Gang Xua, Jun-Ha Yong and Jean- Claude Paul, Computng the mnmum dstane between two Bézer urves, Journal of Computatonal and Appled Mathemats, Vol. 9, 9, pp [] H. Alt, C. Knauer, and C. Wenk, Mathng polygonal urves wth respet to the Frehet dstane, In: Pro. 8th Int. Symp. Theoret. Aspets Comput. S., STACS (A. Ferrera and H. Rehel, eds., Let. Notes Comp. S., vol., Sprnger-Verlag,, pp [3] M. Cu, J. Feman, J. Hu, P. Wonka and A. Razdan, Curve mathng for open D urves, Pattern Reognton Letters, Vol. 3, 9, pp. [4] R. Srraghavendra, Karthk K. and Chranjb Bhattaharyya, Fr ehet Dstane based Approah for Searhng Onlne Handwrtten Douments, Nnth Internatonal Conferene on Doument Analyss and Reognton (ICDAR, Vol., 7 pp ISSN: Issue 5, Volume 9, May

Color Texture Classification using Modified Local Binary Patterns based on Intensity and Color Information

Color Texture Classification using Modified Local Binary Patterns based on Intensity and Color Information Color Texture Classfaton usng Modfed Loal Bnary Patterns based on Intensty and Color Informaton Shvashankar S. Department of Computer Sene Karnatak Unversty, Dharwad-580003 Karnataka,Inda shvashankars@kud.a.n

More information

LOCAL BINARY PATTERNS AND ITS VARIANTS FOR FACE RECOGNITION

LOCAL BINARY PATTERNS AND ITS VARIANTS FOR FACE RECOGNITION IEEE-Internatonal Conferene on Reent Trends n Informaton Tehnology, ICRTIT 211 MIT, Anna Unversty, Chenna. June 3-5, 211 LOCAL BINARY PATTERNS AND ITS VARIANTS FOR FACE RECOGNITION K.Meena #1, Dr.A.Suruland

More information

Cluster ( Vehicle Example. Cluster analysis ( Terminology. Vehicle Clusters. Why cluster?

Cluster (  Vehicle Example. Cluster analysis (  Terminology. Vehicle Clusters. Why cluster? Why luster? referene funton R R Although R and R both somewhat orrelated wth the referene funton, they are unorrelated wth eah other Cluster (www.m-w.om) A number of smlar ndvduals that our together as

More information

Performance Evaluation of TreeQ and LVQ Classifiers for Music Information Retrieval

Performance Evaluation of TreeQ and LVQ Classifiers for Music Information Retrieval Performane Evaluaton of TreeQ and LVQ Classfers for Mus Informaton Retreval Matna Charam, Ram Halloush, Sofa Tsekerdou Athens Informaton Tehnology (AIT) 0.8 km Markopoulo Ave. GR - 19002 Peana, Athens,

More information

Pattern Classification: An Improvement Using Combination of VQ and PCA Based Techniques

Pattern Classification: An Improvement Using Combination of VQ and PCA Based Techniques Ameran Journal of Appled Senes (0): 445-455, 005 ISSN 546-939 005 Sene Publatons Pattern Classfaton: An Improvement Usng Combnaton of VQ and PCA Based Tehnques Alok Sharma, Kuldp K. Palwal and Godfrey

More information

Avatar Face Recognition using Wavelet Transform and Hierarchical Multi-scale LBP

Avatar Face Recognition using Wavelet Transform and Hierarchical Multi-scale LBP 2011 10th Internatonal Conferene on Mahne Learnng and Applatons Avatar Fae Reognton usng Wavelet Transform and Herarhal Mult-sale LBP Abdallah A. Mohamed, Darryl D Souza, Naouel Bal and Roman V. Yampolsky

More information

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Gabor-Filtering-Based Completed Local Binary Patterns for Land-Use Scene Classification

Gabor-Filtering-Based Completed Local Binary Patterns for Land-Use Scene Classification Gabor-Flterng-Based Completed Loal Bnary Patterns for Land-Use Sene Classfaton Chen Chen 1, Lbng Zhou 2,*, Janzhong Guo 1,2, We L 3, Hongjun Su 4, Fangda Guo 5 1 Department of Eletral Engneerng, Unversty

More information

Boosting Weighted Linear Discriminant Analysis

Boosting Weighted Linear Discriminant Analysis . Okada et al. / Internatonal Journal of Advaned Statsts and I&C for Eonoms and Lfe Senes Boostng Weghted Lnear Dsrmnant Analyss azunor Okada, Arturo Flores 2, Marus George Lnguraru 3 Computer Sene Department,

More information

Research on Neural Network Model Based on Subtraction Clustering and Its Applications

Research on Neural Network Model Based on Subtraction Clustering and Its Applications Avalable onlne at www.senedret.om Physs Proeda 5 (01 ) 164 1647 01 Internatonal Conferene on Sold State Deves and Materals Sene Researh on Neural Networ Model Based on Subtraton Clusterng and Its Applatons

More information

Connectivity in Fuzzy Soft graph and its Complement

Connectivity in Fuzzy Soft graph and its Complement IOSR Journal of Mathemats (IOSR-JM) e-issn: 2278-5728, p-issn: 2319-765X. Volume 1 Issue 5 Ver. IV (Sep. - Ot.2016), PP 95-99 www.osrjournals.org Connetvty n Fuzzy Soft graph and ts Complement Shashkala

More information

Fuzzy Modeling for Multi-Label Text Classification Supported by Classification Algorithms

Fuzzy Modeling for Multi-Label Text Classification Supported by Classification Algorithms Journal of Computer Senes Orgnal Researh Paper Fuzzy Modelng for Mult-Label Text Classfaton Supported by Classfaton Algorthms 1 Beatrz Wlges, 2 Gustavo Mateus, 2 Slva Nassar, 2 Renato Cslagh and 3 Rogéro

More information

Link Graph Analysis for Adult Images Classification

Link Graph Analysis for Adult Images Classification Lnk Graph Analyss for Adult Images Classfaton Evgeny Khartonov Insttute of Physs and Tehnology, Yandex LLC 90, 6 Lev Tolstoy st., khartonov@yandex-team.ru Anton Slesarev Insttute of Physs and Tehnology,

More information

Bottom-Up Fuzzy Partitioning in Fuzzy Decision Trees

Bottom-Up Fuzzy Partitioning in Fuzzy Decision Trees Bottom-Up Fuzzy arttonng n Fuzzy eson Trees Maej Fajfer ept. of Mathemats and Computer Sene Unversty of Mssour St. Lous St. Lous, Mssour 63121 maejf@me.pl Cezary Z. Janow ept. of Mathemats and Computer

More information

Adaptive Class Preserving Representation for Image Classification

Adaptive Class Preserving Representation for Image Classification Adaptve Class Preservng Representaton for Image Classfaton Jan-Xun M,, Qankun Fu,, Wesheng L, Chongqng Key Laboratory of Computatonal Intellgene, Chongqng Unversty of Posts and eleommunatons, Chongqng,

More information

Multilabel Classification with Meta-level Features

Multilabel Classification with Meta-level Features Multlabel Classfaton wth Meta-level Features Sddharth Gopal Carnege Mellon Unversty Pttsburgh PA 523 sgopal@andrew.mu.edu Ymng Yang Carnege Mellon Unversty Pttsburgh PA 523 ymng@s.mu.edu ABSTRACT Effetve

More information

FULLY AUTOMATIC IMAGE-BASED REGISTRATION OF UNORGANIZED TLS DATA

FULLY AUTOMATIC IMAGE-BASED REGISTRATION OF UNORGANIZED TLS DATA FULLY AUTOMATIC IMAGE-BASED REGISTRATION OF UNORGANIZED TLS DATA Martn Wenmann, Bors Jutz Insttute of Photogrammetry and Remote Sensng, Karlsruhe Insttute of Tehnology (KIT) Kaserstr. 12, 76128 Karlsruhe,

More information

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram Shape Representaton Robust to the Sketchng Order Usng Dstance Map and Drecton Hstogram Department of Computer Scence Yonse Unversty Kwon Yun CONTENTS Revew Topc Proposed Method System Overvew Sketch Normalzaton

More information

Steganalysis of DCT-Embedding Based Adaptive Steganography and YASS

Steganalysis of DCT-Embedding Based Adaptive Steganography and YASS Steganalyss of DCT-Embeddng Based Adaptve Steganography and YASS Qngzhong Lu Department of Computer Sene Sam Houston State Unversty Huntsvlle, TX 77341, U.S.A. lu@shsu.edu ABSTRACT Reently well-desgned

More information

Bit-level Arithmetic Optimization for Carry-Save Additions

Bit-level Arithmetic Optimization for Carry-Save Additions Bt-leel Arthmet Optmzaton for Carry-Sae s Ke-Yong Khoo, Zhan Yu and Alan N. Wllson, Jr. Integrated Cruts and Systems Laboratory Unersty of Calforna, Los Angeles, CA 995 khoo, zhanyu, wllson @sl.ula.edu

More information

An Adaptive Filter Based on Wavelet Packet Decomposition in Motor Imagery Classification

An Adaptive Filter Based on Wavelet Packet Decomposition in Motor Imagery Classification An Adaptve Flter Based on Wavelet Paket Deomposton n Motor Imagery Classfaton J. Payat, R. Mt, T. Chusak, and N. Sugno Abstrat Bran-Computer Interfae (BCI) s a system that translates bran waves nto eletral

More information

Progressive scan conversion based on edge-dependent interpolation using fuzzy logic

Progressive scan conversion based on edge-dependent interpolation using fuzzy logic Progressve san onverson based on edge-dependent nterpolaton usng fuzzy log P. Brox brox@mse.nm.es I. Baturone lum@mse.nm.es Insttuto de Mroeletróna de Sevlla, Centro Naonal de Mroeletróna Avda. Rena Meredes

More information

DETECTING AND ANALYZING CORROSION SPOTS ON THE HULL OF LARGE MARINE VESSELS USING COLORED 3D LIDAR POINT CLOUDS

DETECTING AND ANALYZING CORROSION SPOTS ON THE HULL OF LARGE MARINE VESSELS USING COLORED 3D LIDAR POINT CLOUDS ISPRS Annals of the Photogrammetry, Remote Sensng and Spatal Informaton Senes, Volume III-3, 2016 XXIII ISPRS Congress, 12 19 July 2016, Prague, Czeh Republ DETECTING AND ANALYZING CORROSION SPOTS ON THE

More information

MULTIPLE OBJECT DETECTION AND TRACKING IN SONAR MOVIES USING AN IMPROVED TEMPORAL DIFFERENCING APPROACH AND TEXTURE ANALYSIS

MULTIPLE OBJECT DETECTION AND TRACKING IN SONAR MOVIES USING AN IMPROVED TEMPORAL DIFFERENCING APPROACH AND TEXTURE ANALYSIS U.P.B. S. Bull., Seres A, Vol. 74, Iss. 2, 2012 ISSN 1223-7027 MULTIPLE OBJECT DETECTION AND TRACKING IN SONAR MOVIES USING AN IMPROVED TEMPORAL DIFFERENCING APPROACH AND TEXTURE ANALYSIS Tudor BARBU 1

More information

Matrix-Matrix Multiplication Using Systolic Array Architecture in Bluespec

Matrix-Matrix Multiplication Using Systolic Array Architecture in Bluespec Matrx-Matrx Multplaton Usng Systol Array Arhteture n Bluespe Team SegFault Chatanya Peddawad (EEB096), Aman Goel (EEB087), heera B (EEB090) Ot. 25, 205 Theoretal Bakground. Matrx-Matrx Multplaton on Hardware

More information

Interval uncertain optimization of structures using Chebyshev meta-models

Interval uncertain optimization of structures using Chebyshev meta-models 0 th World Congress on Strutural and Multdsplnary Optmzaton May 9-24, 203, Orlando, Florda, USA Interval unertan optmzaton of strutures usng Chebyshev meta-models Jngla Wu, Zhen Luo, Nong Zhang (Tmes New

More information

Pixel-Based Texture Classification of Tissues in Computed Tomography

Pixel-Based Texture Classification of Tissues in Computed Tomography Pxel-Based Texture Classfaton of Tssues n Computed Tomography Ruhaneewan Susomboon, Danela Stan Rau, Jaob Furst Intellgent ultmeda Proessng Laboratory Shool of Computer Sene, Teleommunatons, and Informaton

More information

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Optimal shape and location of piezoelectric materials for topology optimization of flextensional actuators

Optimal shape and location of piezoelectric materials for topology optimization of flextensional actuators Optmal shape and loaton of pezoeletr materals for topology optmzaton of flextensonal atuators ng L 1 Xueme Xn 2 Noboru Kkuh 1 Kazuhro Satou 1 1 Department of Mehanal Engneerng, Unversty of Mhgan, Ann Arbor,

More information

Measurement and Calibration of High Accuracy Spherical Joints

Measurement and Calibration of High Accuracy Spherical Joints 1. Introduton easurement and Calbraton of Hgh Auray Spheral Jonts Ale Robertson, Adam Rzepnewsk, Alexander Sloum assahusetts Insttute of Tehnolog Cambrdge, A Hgh auray robot manpulators are requred for

More information

Session 4.2. Switching planning. Switching/Routing planning

Session 4.2. Switching planning. Switching/Routing planning ITU Semnar Warsaw Poland 6-0 Otober 2003 Sesson 4.2 Swthng/Routng plannng Network Plannng Strategy for evolvng Network Arhtetures Sesson 4.2- Swthng plannng Loaton problem : Optmal plaement of exhanges

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

FUZZY SEGMENTATION IN IMAGE PROCESSING

FUZZY SEGMENTATION IN IMAGE PROCESSING FUZZY SEGMENTATION IN IMAGE PROESSING uevas J. Er,, Zaldívar N. Danel,, Roas Raúl Free Unverstät Berln, Insttut für Inforat Tausstr. 9, D-495 Berln, Gerany. Tel. 0049-030-8385485, Fax. 0049-030-8387509

More information

AUTOMATICALLY MULTIPLE FEATURES DETECTION OF FACE SKETCH BASED ON MAXIMUM LINE GRADIENT

AUTOMATICALLY MULTIPLE FEATURES DETECTION OF FACE SKETCH BASED ON MAXIMUM LINE GRADIENT AUTOMATICALLY MULTIPLE FEATURES DETECTION OF FACE SKETCH BASED ON MAXIMUM LINE GRADIENT Arf Muntasa, Mohamad Harad, Maurdh Her Purnomo 3,,3 Eletral Engneerng Department, Insttut Teknolog Sepuluh Nopember,

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

More information

Improved Accurate Extrinsic Calibration Algorithm of Camera and Two-dimensional Laser Scanner

Improved Accurate Extrinsic Calibration Algorithm of Camera and Two-dimensional Laser Scanner JOURNAL OF MULTIMEDIA, VOL. 8, NO. 6, DECEMBER 013 777 Improved Aurate Extrns Calbraton Algorthm of Camera and Two-dmensonal Laser Sanner Janle Kong, Le Yan*, Jnhao Lu, Qngqng Huang, and Xaokang Dng College

More information

The Simulation of Electromagnetic Suspension System Based on the Finite Element Analysis

The Simulation of Electromagnetic Suspension System Based on the Finite Element Analysis 308 JOURNAL OF COMPUTERS, VOL. 8, NO., FEBRUARY 03 The Smulaton of Suspenson System Based on the Fnte Element Analyss Zhengfeng Mng Shool of Eletron & Mahanal Engneerng, Xdan Unversty, X an, Chna Emal:

More information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

ON THE USE OF THE SIFT TRANSFORM TO SELF-LOCATE AND POSITION EYE-IN-HAND MANIPULATORS USING VISUAL CONTROL

ON THE USE OF THE SIFT TRANSFORM TO SELF-LOCATE AND POSITION EYE-IN-HAND MANIPULATORS USING VISUAL CONTROL XVIII Congresso Braslero de Automáta / a 6-setembro-00, Bonto-MS ON THE USE OF THE SIFT TRANSFORM TO SELF-LOCATE AND POSITION EYE-IN-HAND MANIPULATORS USING VISUAL CONTROL ILANA NIGRI, RAUL Q. FEITOSA

More information

A Fast Way to Produce Optimal Fixed-Depth Decision Trees

A Fast Way to Produce Optimal Fixed-Depth Decision Trees A Fast Way to Produe Optmal Fxed-Depth Deson Trees Alreza Farhangfar, Russell Grener and Martn Znkevh Dept of Computng Sene Unversty of Alberta Edmonton, Alberta T6G 2E8 Canada {farhang, grener, maz}@s.ualberta.a

More information

AVideoStabilizationMethodbasedonInterFrameImageMatchingScore

AVideoStabilizationMethodbasedonInterFrameImageMatchingScore Global Journal of Computer Sene and Tehnology: F Graphs & vson Volume 7 Issue Verson.0 Year 207 Type: Double Blnd Peer Revewed Internatonal Researh Journal Publsher: Global Journals In. (USA) Onlne ISSN:

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

A Flexible Solution for Modeling and Tracking Generic Dynamic 3D Environments*

A Flexible Solution for Modeling and Tracking Generic Dynamic 3D Environments* A Flexble Soluton for Modelng and Trang Gener Dynam 3D Envronments* Radu Danesu, Member, IEEE, and Sergu Nedevsh, Member, IEEE Abstrat The traff envronment s a dynam and omplex 3D sene, whh needs aurate

More information

A Toolbox for Easily Calibrating Omnidirectional Cameras

A Toolbox for Easily Calibrating Omnidirectional Cameras A oolbox for Easly Calbratng Omndretonal Cameras Davde Saramuzza, Agostno Martnell, Roland Segwart Autonomous Systems ab Swss Federal Insttute of ehnology Zurh EH) CH-89, Zurh, Swtzerland {davdesaramuzza,

More information

3D Scene Reconstruction System from Multiple Synchronized Video Images

3D Scene Reconstruction System from Multiple Synchronized Video Images 3D Sene Reonstruton Sstem from Multple Snhronzed Vdeo Images aewoo Han 1, Juho Lee 2, Hung S. Yang 3 AIM Lab., EE/CS Dept., KAIS 1,2,3 373-1, Guseong-dong, Yuseong-gu, Daejon, Republ of Korea { bluebrd

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

REGISTRATION OF TERRESTRIAL LASER SCANNER DATA USING IMAGERY INTRODUCTION

REGISTRATION OF TERRESTRIAL LASER SCANNER DATA USING IMAGERY INTRODUCTION EGISTATION OF TEESTIAL LASE SCANNE DATA USING IMAGEY Khall Al-Manasr, Ph.D student Clve S. Fraser, Professor Department of Geomats Unversty of Melbourne Vtora 3010 Australa k.al-manasr@pgrad.unmelb.edu.au.fraser@unmelb.edu.au

More information

Multi-scale and Discriminative Part Detectors Based Features for Multi-label Image Classification

Multi-scale and Discriminative Part Detectors Based Features for Multi-label Image Classification Proeedngs of the wenty-seventh Internatonal Jont Conferene on Artfal Intellgene (IJCAI-8) Mult-sale and Dsrmnatve Part Detetors Based Features for Mult-lael Image Classfaton Gong Cheng, Deheng Gao, Yang

More information

A MPAA-Based Iterative Clustering Algorithm Augmented by Nearest Neighbors Search for Time-Series Data Streams

A MPAA-Based Iterative Clustering Algorithm Augmented by Nearest Neighbors Search for Time-Series Data Streams A MPAA-Based Iteratve Clusterng Algorthm Augmented by Nearest Neghbors Searh for Tme-Seres Data Streams Jessa Ln 1, Mha Vlahos 1, Eamonn Keogh 1, Dmtros Gunopulos 1, Janwe Lu 2, Shouan Yu 2, and Jan Le

More information

Multiscale Heterogeneous Modeling with Surfacelets

Multiscale Heterogeneous Modeling with Surfacelets 759 Multsale Heterogeneous Modelng wth Surfaelets Yan Wang 1 and Davd W. Rosen 2 1 Georga Insttute of Tehnology, yan.wang@me.gateh.edu 2 Georga Insttute of Tehnology, davd.rosen@me.gateh.edu ABSTRACT Computatonal

More information

Visual Curvature. 1. Introduction. y C. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), June 2007

Visual Curvature. 1. Introduction. y C. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), June 2007 IEEE onf. on omputer Vson and Pattern Recognton (VPR June 7 Vsual urvature HaRong Lu, Longn Jan Lateck, WenYu Lu, Xang Ba HuaZhong Unversty of Scence and Technology, P.R. hna Temple Unversty, US lhrbss@gmal.com,

More information

USING GRAPHING SKILLS

USING GRAPHING SKILLS Name: BOLOGY: Date: _ Class: USNG GRAPHNG SKLLS NTRODUCTON: Recorded data can be plotted on a graph. A graph s a pctoral representaton of nformaton recorded n a data table. t s used to show a relatonshp

More information

Scale Selective Extended Local Binary Pattern For Texture Classification

Scale Selective Extended Local Binary Pattern For Texture Classification Scale Selectve Extended Local Bnary Pattern For Texture Classfcaton Yutng Hu, Zhlng Long, and Ghassan AlRegb Multmeda & Sensors Lab (MSL) Georga Insttute of Technology 03/09/017 Outlne Texture Representaton

More information

Some Advanced SPC Tools 1. Cumulative Sum Control (Cusum) Chart For the data shown in Table 9-1, the x chart can be generated.

Some Advanced SPC Tools 1. Cumulative Sum Control (Cusum) Chart For the data shown in Table 9-1, the x chart can be generated. Some Advanced SP Tools 1. umulatve Sum ontrol (usum) hart For the data shown n Table 9-1, the x chart can be generated. However, the shft taken place at sample #21 s not apparent. 92 For ths set samples,

More information

International Journal of Pharma and Bio Sciences HYBRID CLUSTERING ALGORITHM USING POSSIBILISTIC ROUGH C-MEANS ABSTRACT

International Journal of Pharma and Bio Sciences HYBRID CLUSTERING ALGORITHM USING POSSIBILISTIC ROUGH C-MEANS ABSTRACT Int J Pharm Bo S 205 Ot; 6(4): (B) 799-80 Researh Artle Botehnology Internatonal Journal of Pharma and Bo Senes ISSN 0975-6299 HYBRID CLUSTERING ALGORITHM USING POSSIBILISTIC ROUGH C-MEANS *ANURADHA J,

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data Malaysan Journal of Mathematcal Scences 11(S) Aprl : 35 46 (2017) Specal Issue: The 2nd Internatonal Conference and Workshop on Mathematcal Analyss (ICWOMA 2016) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES

More information

A Robust Algorithm for Text Detection in Color Images

A Robust Algorithm for Text Detection in Color Images A Robust Algorthm for Tet Deteton n Color Images Yangng LIU Satosh GOTO Takesh IKENAGA Abstrat Tet deteton n olor mages has beome an atve researh area sne reent deades. In ths paper we present a novel

More information

A Model-Based Approach for Automated Feature Extraction in Fundus Images

A Model-Based Approach for Automated Feature Extraction in Fundus Images A Model-Based Approah for Automated Feature Extraton n Fundus Images Huq L Shool of Computng Natonal Unversty of Sngapore dslhq@nus.edu.sg Opas Chutatape Shool of Eletral and Eletron Engneerng Nanyang

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications 14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of

More information

Clustering incomplete data using kernel-based fuzzy c-means algorithm

Clustering incomplete data using kernel-based fuzzy c-means algorithm Clusterng noplete data usng ernel-based fuzzy -eans algorth Dao-Qang Zhang *, Song-Can Chen Departent of Coputer Sene and Engneerng, Nanjng Unversty of Aeronauts and Astronauts, Nanjng, 210016, People

More information

Semi-analytic Evaluation of Quality of Service Parameters in Multihop Networks

Semi-analytic Evaluation of Quality of Service Parameters in Multihop Networks U J.T. (4): -4 (pr. 8) Sem-analyt Evaluaton of Qualty of Serve arameters n Multhop etworks Dobr tanassov Batovsk Faulty of Sene and Tehnology, ssumpton Unversty, Bangkok, Thaland bstrat

More information

Computing Cloud Cover Fraction in Satellite Images using Deep Extreme Learning Machine

Computing Cloud Cover Fraction in Satellite Images using Deep Extreme Learning Machine Computng Cloud Cover Fraton n Satellte Images usng Deep Extreme Learnng Mahne L-guo WENG, We-bn KONG, Mn XIA College of Informaton and Control, Nanjng Unversty of Informaton Sene & Tehnology, Nanjng Jangsu

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between

More information

A Novel Dynamic and Scalable Caching Algorithm of Proxy Server for Multimedia Objects

A Novel Dynamic and Scalable Caching Algorithm of Proxy Server for Multimedia Objects Journal of VLSI Sgnal Proessng 2007 * 2007 Sprnger Sene + Busness Meda, LLC. Manufatured n The Unted States. DOI: 10.1007/s11265-006-0024-7 A Novel Dynam and Salable Cahng Algorthm of Proxy Server for

More information

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval Fuzzy -Means Intalzed by Fxed Threshold lusterng for Improvng Image Retreval NAWARA HANSIRI, SIRIPORN SUPRATID,HOM KIMPAN 3 Faculty of Informaton Technology Rangst Unversty Muang-Ake, Paholyotn Road, Patumtan,

More information

An Approach in Coloring Semi-Regular Tilings on the Hyperbolic Plane

An Approach in Coloring Semi-Regular Tilings on the Hyperbolic Plane An Approach n Colorng Sem-Regular Tlngs on the Hyperbolc Plane Ma Louse Antonette N De Las Peñas, mlp@mathscmathadmueduph Glenn R Lago, glago@yahoocom Math Department, Ateneo de Manla Unversty, Loyola

More information

Machine Learning 9. week

Machine Learning 9. week Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below

More information

3D vector computer graphics

3D vector computer graphics 3D vector computer graphcs Paolo Varagnolo: freelance engneer Padova Aprl 2016 Prvate Practce ----------------------------------- 1. Introducton Vector 3D model representaton n computer graphcs requres

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

On the End-to-end Call Acceptance and the Possibility of Deterministic QoS Guarantees in Ad hoc Wireless Networks

On the End-to-end Call Acceptance and the Possibility of Deterministic QoS Guarantees in Ad hoc Wireless Networks On the End-to-end Call Aeptane and the Possblty of Determnst QoS Guarantees n Ad ho Wreless Networks S. Srram T. heemarjuna Reddy Dept. of Computer Sene Dept. of Computer Sene and Engneerng Unversty of

More information

Clustering Algorithm of Similarity Segmentation based on Point Sorting

Clustering Algorithm of Similarity Segmentation based on Point Sorting Internatonal onference on Logstcs Engneerng, Management and omputer Scence (LEMS 2015) lusterng Algorthm of Smlarty Segmentaton based on Pont Sortng Hanbng L, Yan Wang*, Lan Huang, Mngda L, Yng Sun, Hanyuan

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

y and the total sum of

y and the total sum of Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton

More information

Recognizing Faces. Outline

Recognizing Faces. Outline Recognzng Faces Drk Colbry Outlne Introducton and Motvaton Defnng a feature vector Prncpal Component Analyss Lnear Dscrmnate Analyss !"" #$""% http://www.nfotech.oulu.f/annual/2004 + &'()*) '+)* 2 ! &

More information

arxiv: v3 [cs.cv] 31 Oct 2016

arxiv: v3 [cs.cv] 31 Oct 2016 Unversal Correspondene Network Chrstopher B. Choy Stanford Unversty hrshoy@a.stanford.edu JunYoung Gwak Stanford Unversty jgwak@a.stanford.edu Slvo Savarese Stanford Unversty sslvo@stanford.edu arxv:1606.03558v3

More information

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints TPL-ware Dsplacement-drven Detaled Placement Refnement wth Colorng Constrants Tao Ln Iowa State Unversty tln@astate.edu Chrs Chu Iowa State Unversty cnchu@astate.edu BSTRCT To mnmze the effect of process

More information

Evaluation of Segmentation in Magnetic Resonance Images Using k-means and Fuzzy c-means Clustering Algorithms

Evaluation of Segmentation in Magnetic Resonance Images Using k-means and Fuzzy c-means Clustering Algorithms ELEKTROTEHIŠKI VESTIK 79(3): 129-134, 2011 EGLISH EDITIO Evaluaton of Segmentaton n Magnet Resonane Images Usng k-means and Fuzzy -Means Clusterng Algorthms Tomaž Fnkšt Unverza v Lublan, Fakulteta za stronštvo,

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning Computer Anmaton and Vsualsaton Lecture 4. Rggng / Sknnng Taku Komura Overvew Sknnng / Rggng Background knowledge Lnear Blendng How to decde weghts? Example-based Method Anatomcal models Sknnng Assume

More information

Professional competences training path for an e-commerce major, based on the ISM method

Professional competences training path for an e-commerce major, based on the ISM method World Transactons on Engneerng and Technology Educaton Vol.14, No.4, 2016 2016 WIETE Professonal competences tranng path for an e-commerce maor, based on the ISM method Ru Wang, Pn Peng, L-gang Lu & Lng

More information

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

Performance Evaluation of Information Retrieval Systems

Performance Evaluation of Information Retrieval Systems Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence

More information

Minimize Congestion for Random-Walks in Networks via Local Adaptive Congestion Control

Minimize Congestion for Random-Walks in Networks via Local Adaptive Congestion Control Journal of Communatons Vol. 11, No. 6, June 2016 Mnmze Congeston for Random-Walks n Networks va Loal Adaptve Congeston Control Yang Lu, Y Shen, and Le Dng College of Informaton Sene and Tehnology, Nanjng

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information